I have a background in electronics and I am comfortable with a programming language. I want to make and sell my own product on the market. there is a lot demand for Vision system in my country

I am sharing my planning details

Product details

• To check and verify OCR, Barcode, Pharmacode, 2D code on Label. These OCR contains the detail of the product like Batch no., M.R.P., Manufacturing date, expiry date etc.
• Track and Trace System to get the data of the product for Pharma product.

TECHNICAL SPECIFICATION:

Hardware

• Industrial PC
• Vision Camera
• Sensor
• Air/Piston based Rejection System
• Conveyor belt

Software

• There is software in the PC should inspects a particular image of the product and check that inspection criterion are fulfilled or not
  
• If product is failed to fulfill pre-defined inspection criteria, Software generates Rejection signal and transfers that signal to Rejection system so that product is rejected and stores values or images in report file which can be rejected
  
• Rejection System may be Air based or Piston based, depending upon weight and dimension of the product. On receiving rejection signal, rejection system rejects/sorts particular defective product from the production line

I know very well it's not easy to make and it may take one or two years time I have a lot of time to make my own product and I am ready to invest my time and effort.

I need a guideline to make this automated inspection systems. I know the requirements but I do not understand how to start. I would appreciate your thoughts and assistance

note : feel free to ask any doubts

 

My experience with AOI (Automated Optical Inspection) is that when a PCB fails AOI the machine stops and holds the product in place within the AOI machine. An operator reviews each reject point and determines whether the reject is correctly rejected or not. After human review the PCB is either sent to the next phase of assembly or it's redirected for rework. The operator identifies each legitimate point for rework.

AOI is best suited for identifying parts with numbers and polarized parts. I've seen it used to evaluate solder, but that's less accurate as low solder may still be acceptable per IPC standards. Simply kicking out a board that has a thousand components on it without verifying the legitimacy of reject points AND without identifying the legitimate points for rework means an inspector is going to have to visually inspect the entire board, which also means component identification and orientation. And many SMT components are so small that at most they can be printed with a code such as Q25L, which doesn't mean anything in itself. You have to look up the manufacture, what the component is supposed to be, then verify whether the part marking matches. It's slow and time consuming. Granted, AOI can be a lot quicker, but I've seen AOI make mistakes. Especially when someone programs in 10 acceptable pictures for a single component. The system is only as good as the programmer. And nearly always programming starts with a known good board. The AOI takes a picture at every location. As you know, a resistor can be installed either way, so the label on the resistor that says 103 can be up-side-down and can appear as E01. Which is sometimes code for a different component.

Engineering an AOI system can be quite the task. From time to engineering, to programming and display images, you're looking at a lot of time and money to develop one. That's why they're so expensive. Far more costly to build one than to buy one.

Good luck.

 

I have designed a lot of automated and semi-automated test machines, and you certainly have an optomistic goal in mind. Your very first challenge will be to create software to read accurately. THAT is quite a task in itself. Then the part about verifying will be another huge task, all in the software and data gathering, because you will need to be able to get data about products that is both accurate and quick. Probably that will be what stops the project.

 

The system is only as good as the programmer. And nearly always programming starts with a known good board.

Good luck.

Your very first challenge will be to create software to read accurately. THAT is quite a task in itself. Then the part about verifying will be another huge task, all in the software and data gathering, because you will need to be able to get data about products that is both accurate and quick. .

Thank you so much for taking interest in my thread. I think programming should start later. I would first know about the hardware which parts I would need.

My first doubt is, how do I decide which parts I need to make a project. There are different part's available in the market with different specification.

How to choose the right one that would be suitable for my project?

 

The type of product is not mentioned, and so any thoughts about what is required to inspect it is just a wild guess. AND, given the amount of data that would be analysed, probably a printer to produce a written report would be fast enough, and likewise, hand loading and unloading should be fast enough.

 

The type of product is not mentioned, and so any thoughts about what is required to inspect it is just a wild guess

Product details

• The system should be used for barcode, batch code verification on the labels
• To check and verify OCR, Barcode, Pharmacode, 2D code on Label. These OCR contains the detail of the product like Batch no., M.R.P., Manufacturing date, expiry date etc.
• Track and Trace System to get the data of the product for Pharma product.

Here is something

 

The real challenge will be the software that first must interpret the images and the resulting data, and then do a comparison with stored data to compare the data. If it is only checking one type of product at a time that will not be so very difficult.
And still, while the descriptions of the actions are described there is a whole lot of information missing. To verify anything there must be some reference to verify against. And so the whole system will depend on having all of the reference data on hand to do the comparison. The mechanical details are the much simpler part of the project.
So here again we have a post asking for advice but not providing the needed information. And without adequate understanding all of the advice is based on speculation and guesses.

 

Sounds more like barcode readers than AOI.

Now I'm wondering why you'd reject a package because of its barcode?

 

Sounds more like barcode readers than AOI.

Now I'm wondering why you'd reject a package because of its barcode?

I am guessing that this is intended to be a system to detect counterfeit products, a large problem in some areas, and a serious problem wherever it occurs. But I am also considering that the software and the resources needed are going to be rather complex.

I am hoping that somebody can create a method of detecting spoofed caller ID signatures, as they have become a nasty plague recently.

 

I am hoping that somebody can create a method of detecting spoofed caller ID signatures, as they have become a nasty plague recently.

I agree. Got a call from the "United States Treasury". I said to the guy who sounded middle eastern, "You don't sound like an American." He paused a minute, then said "I'm from Canada." So I asked "What do you want?" He said "Money." I said "Well, you're not getting any from me." He hung up. What was funny about the call was that I was in Wisconsin when it happened. I'm from Utah, and the caller ID said it was a call from within Utah. I get lots of local calls from scammers. I know they're not local.

I even get e-mail with my own personal e-mail address as the "Senders" address. I KNOW that's spoofing. Counterfeit IS a big problem. Working in Wisconsin at a defense contractors site, we had a lot of training on how to recognize counterfeit parts. Labeling - well, I don't readily know how you'd be detecting counterfeits, but that's an area I haven't any experience.

Grocery stores have bar-code readers. Is that sort of what the TS is looking for? I know Amazon has a system to direct packaging through a complex conveyor system so that your parts get shipped to you without having to go to the North Pole first. That's some sort of optical reader system, and even cell phones have optical recognition with that new-fangled square information box. What's that called? A fiducial? Probably not. It's important to fully understand what the TS is trying to accomplish. That's the best way we can help.

 

Detection of counterfeit products is vital in many areas, especially in medicines. And there is so much money to be made selling them that a whole lot of "bad guys" are in the business.. Of course there are also a fair number of products that are only identified by the assorted codes, and so a machine to provide plain language description labels could be quite useful.
But a device to detect and intercept spoofed caller ID calls is a product that I would certainly buy, since I need to answer every call because they may be new clients or customers. And it does not appear that any amount of verbal abuse deters these jerks from calling again.

 

So here again we have a post asking for advice but not providing the needed information..

What the information you need ?

I think I have explained all the requirements. I have to make planning according to given requirements

The first step I need a smart camera that can take the image of the barcode on labels then I have to make the program to extracting text from images based on OCR OCV

Which camera choose for my project?

 

The camera will depend on the size and height of the objects and the sort speed.

And that speed will depend on your software.

How fast can you do the comparison needed(longest case) and output a signal?

 

The camera will depend on the size and height of the objects and the sort speed.

And that speed will depend on your software.

How fast can you do the comparison needed(longest case) and output a signal?

object size
height = 85 mm
length = 190 mm

I want the camera should be able to check 500 labels /minute

 

The camera will depend on the size and height of the objects and the sort speed.

And that speed will depend on your software.

How fast can you do the comparison needed(longest case) and output a signal?

Hello BR-549

What do you think about the specification given in the link http://softwareservices.ptgrey.com/BFS-U3-200S6/latest/Model/spec.html

Will you suggest me a good vision camera ?

 

No, I couldn't do that. I have no idea of what is required. Because I have no idea of what you are doing.

A general description of sorting objects on a conveyor via bar code,........means nothing for hardware assessment.

The hardware and the method WILL depend on the speed of your software and it also depends on the reject criteria.

This is not an easy task at high speed. Preparing the sort can also help. If the label is in same location and orientation.....it can speed things up.

This job takes the talent of a systems integrator.

 

I thought that I saw a rate of 500 packages per minute, although I do not see it today. That is quite impressive and will need HUGE amounts of fast computer processing power. Far better to run several tracks in parallel and use slower processing. Image recognition alone will be a big deal effort.. So the hardware will be costly but simple to design, the software will be the challenging part

 

500 labels a minute. 8.33 labels a second. How much belt length per object? 1 foot?

Lets plan on 9 labels a second. 540 labels and ft per minute. 6.17 MPH. 111 milliseconds between each label.

We need to be able to cleanly reject in that time. Shorter actually. So....maybe an air jet reject. They sort potato chips with air. Sounds like "air pops".

It should be designed for 600 per minute. One could give rejection a length of belt.....to give software more time. That takes more length and more area.

And as already mentioned......lots of sorts have parallel lines. More area and cost. And bandwidth.

The mystery is why the camera image is needed. That was the whole idea of the bar code.

 

Why the system needs a camera is contained in the first post::

Product details

• To check and verify OCR, Barcode, Pharmacode, 2D code on Label. These OCR contains the detail of the product like Batch no., M.R.P., Manufacturing date, expiry date etc
• .So it has a whole lot to check in that 111 milliseconds, and given the nebulous descriptions I doubt that such a machine will come to exist. Perhaps one item per second and multiple lanes. OR multiple reads looking for different things. Sequential checking works well, it just makes keeping track of results a bit more complex. But sequential checking by different systems is certainly one effective way to increase through-put while having more time for analysis. And it may cost less than having much higher powered computers and sensors.

 

500 labels a minute. 8.33 labels a second. How much belt length per object? 1 foot?

Lets plan on 9 labels a second. 540 labels and ft per minute. 6.17 MPH. 111 milliseconds between each label.

Where I worked they used the 'Vision' system on the presses that made terminals. They looked at 4 or more areas on connector terminals as they were being made, to make sure they met the standards. These presses ran at 1300 strokes per minute, or 1300 terminals per minute. When a defect was found they shut the press down within 10 seconds of finding the third bad terminal in a row. These were 25 Ton mechanical stamping presses.